Oceanic processes associated with anomalous eventsin the Indian Ocean with relevance to 1997–1998

Raghu MurtuguddeEarth System Science Interdisciplinary Center, University of Maryland, College Park

Julian P. McCreary Jr.International Pacific Research Center, University of Hawaii, Honolulu

Antonio J. BusalacchiLaboratory for Hydrospheric Processes, NASA Goddard Space Flight Center, Greenbelt, Maryland

Abstract. An anomalous climatic event occurred in the Indian Ocean (IO) region during1997–1998, which coincided with a severe drought in Indonesia and floods in parts ofeastern Africa. Cool sea surface temperature anomalies (SSTAs) were present in theeastern IO along and south of the equator. Beginning in July 1997, warm SSTAs appearedin the western IO, and they peaked in February 1998. An ocean general circulation modelis employed to investigate the dynamic and thermodynamic processes that caused theSSTAs associated with this and other similar IO events. The eastern cooling resulted fromunusually strong upwelling along the equator and Sumatra. The Sumatran upwelling wasforced both locally by the stronger alongshore winds and remotely by equatorial andcoastal Kelvin waves. By the end of 1997, weakening of the winds and the associatedreduction in latent heat loss led to the elimination of the cold SST anomalies in the east.The western warming was initiated by weaker Southwest Monsoon winds and maintainedby enhanced precipitation forcing, which resulted in a barrier layer structure. Analysis ofthe mixed layer temperature equation indicates that a downwelling Rossby wavecontribution was crucial for sustaining the warming into February 1998. It is tempting tosuppose that the 1997 event was related to the El Nino-Southern Oscillation (ENSO)event that took place in the Pacific at the same time. Indeed, weaker IO events occurquite regularly in the control run that evolve similarly to the 1997 event, and they areoften but not always related to ENSO. We speculate that these events represent a naturalmode of oscillation in the IO, which is externally forced by ENSO but also excited byocean-atmosphere interactions internal to the IO.

1. Introduction

An extreme climatic event took place in the Indian Ocean(IO) during 1997 and early 1998. It was associated with thesecond-largest sea surface temperature anomalies (SSTAs) oc-curring in the region during the past half century. Plate 1aillustrates the time development of observed SSTAs along theequator during the event. In the late spring of 1997 a surfacewarming ensued over the western region. (Throughout thispaper, seasons always refer to those in the Northern Hemi-sphere.) It peaked in February 1998, by which time it hadspread across most of the tropical IO.

Strong cooling was evident in the eastern ocean by Septem-ber 1997, and it continued into January 1998. As a conse-quence of these changes, there was an equatorial SST gradientfrom east to west, a reversal of the climatological situation. Thewind stress field was also highly unusual during the fall, witheasterlies along the equator replacing the climatological west-erlies (Plate 2a) and strengthened alongshore winds off

Sumatra. As the western warming continued, anomalous con-vection and rainfall over eastern Africa and the western IndianOcean developed and persisted [Birkett et al., 1999].

It is reasonable to suppose that this remarkable IO event wasrelated to the El Nino-Southern Oscillation (ENSO) event thattook place at the same time in the Pacific. In fact, a number ofsimilar but weaker IO events occurred during the past 40 yearsthat are also related to ENSO (see the discussion in section3.3). On the other hand, the ENSO connection is not exact. Forexample, the largest IO event during the same time intervaltook place in 1961, which is not recognized as an ENSO year[Reverdin et al., 1986].

In this paper we utilize an oceanic general circulation model(OGCM) to investigate the dynamics and thermodynamics ofsuch IO events during the past 40 years, focussing on the 1997extreme event. Specifically, we first identify the basic processesthat caused warm and cold SSTAs during the 1997 event. Thenwe demonstrate that other weaker IO events evolved similarly.This similarity indicates that they represent a natural mode ofoscillation in the IO, one that is easily excited in a variety ofways. We note that it appears to be externally forced by ENSOand speculate that it is also generated by ocean-atmosphereinteractions internal to the IO.

Copyright 2000 by the American Geophysical Union.

Paper number 1999JC900294.0148-0227/00/1999JC900294$09.00

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 105, NO. C2, PAGES 3295–3306, FEBRUARY 15, 2000

3295

2. Model OceanThe ocean model is a reduced gravity primitive equation

OGCM confined to the IO region north of 308S. It consists of19 sigma coordinate layers beneath a variable thickness surfacemixed layer [Chen et al., 1994]. Surface heat fluxes are deter-mined by coupling a simple model of the atmospheric bound-ary layer to the OGCM, and freshwater forcing is included asa natural boundary condition. Model details are presented byMurtugudde et al. [1996, 1998a] and Murtugudde and Busalacchi[1999]. Model salinity, temperature, and layer thicknesses arerelaxed to Levitus [1994] data from 308 to 258S at the southernboundary. For our control run the model is almost alwaysforced by 10 m winds and surface precipitation from NationalCenters Environmental Prediction (NCEP) reanalyses [Kalnayet al., 1996] for the period from January 1958 to March 1998.The exception is for the period from October 1997 to February1998, when the NCEP wind stress was reduced by 20% in theSomali Jet region (508–708E, 08–208N) and also along the equa-tor (from 58S to 58N) after they were found to be too strongcompared to Special Sensor Microwave Imager (SSMI) windproducts. In addition, for the period mentioned above thegustiness parameter, which accounts for latent heat loss due to

high-frequency wind forcing (diurnal to weekly), had to be setto 0.5 m s21 for regions with low wind speeds (see Plate 5),whereas it is normally set to 4.5 m s21 over the whole basin[Murtugudde and Busalacchi, 1998]. The reduced minimumwind speed was important for simulating the dissipation of coldSSTAs in the east, and the reduction in wind stresses allowedaccurate simulation of warming in the west.

Because of a lack of sufficiently long time series of reliablecloudiness and solar radiation data, climatological Earth Ra-diation Budget Experiment (ERBE) radiation and Interna-tional Satellite Cloud Climatology Project (ISCCP) cloudinessare used in the interannual simulations to determine surfaceheat fluxes [also see Murtugudde et al., 1996, 1998a]. Windspeed for the latent and sensible heat fluxes is computed fromthe interannual wind stresses assuming a drag coefficient of0.0015. We also obtained a solution forced by climatologicalprecipitation, wind stresses, and wind speed, and throughoutthe text we refer to it as the “climatological solution.”

It is well known that there are inaccuracies in the availableatmospheric forcing fields. To investigate effects due to theseinaccuracies, we also obtained a number of test experimentsdriven by different products. There are significant differences

Plate 1. Anomolous sea surface temperatures (SSTs) (the contour interval is 0.258C) along the equatorduring 1997–1998 for the (a) Reynolds SST and (b) control run. The model was driven with interannual windsand precipitation but climatological radiation and cloudiness.

MURTUGUDDE ET AL.: 1997–1998 INDIAN OCEAN ANOMALOUS EVENTS3296

among the solutions. For example, in a run forced by FloridaState University (FSU), winds for 1997–1998 SSTAs werecooler by 0.68C in the western ocean, and off Sumatra theywere colder by as much as 38C because of much strongercoastal upwelling there. When the ISCCP radiation climatol-ogy was employed instead of ERBE data, the eastern equato-

rial IO was warmer by ;0.88C during April 1998 [also seeMurtugudde and Busalacchi, 1999]. With forcing by interannualprecipitation (both NCEP and Xie and Arkin [1996] data pro-duced similar results), SSTAs were cooler by 0.68C in theeastern ocean and warmer by 0.58C in the west: lower precip-itation in the east during fall 1997 facilitates easier erosion of

Plate 2. Anomalous fields during October–December 1997 showing (a) National Centers for EnvironmentalPrediction (NCEP) wind stress vectors (dynes cm22) overlying wind speed anomalies (m s21) and (b) surfacecurrent vectors (m s21) over SST (degrees Celsius) for the control run. The thick line corresponds to the 08Ccontour.

3297MURTUGUDDE ET AL.: 1997–1998 INDIAN OCEAN ANOMALOUS EVENTS

the barrier layer and enhanced cooling; higher precipitation inthe west leads to a freshwater lens, a shallower mixed layer,and warmer SSTs. Despite these differences the general pat-tern of the responses was consistent among all the solutions.Given this consistency, we believe that our control run is suf-ficiently accurate to capture the dominant physics of the event.

3. ResultsIn this section we first describe the 1997–1998 event in our

control run (section 3.1) and discuss the dynamic and thermo-dynamic processes that account for its prominent SSTAs (sec-tion 3.2). Then we note that weaker events occur regularly andhave a similar structure to the 1997 event (section 3.3).

3.1. Description

3.1.1. SST anomalies. Observed (Plate 1a) and modeled(Plate 1b) SSTAs along the equator for 1997–1998 are con-trasted. The modeled SSTA is able to reproduce the basin-scale features of the event (i.e., eastern cooling and westernwarming), albeit with some notable differences from the ob-served SSTA. In the east, coldest SSTAs occur in Novemberclose to the eastern boundary, consistent with the observations,but the minimum is warmer by 0.758C. In the west the model’sSSTAs are slightly warmer in the fall but cooler in the winter,especially during January–February 1998. These discrepanciesmay be partly due to inaccuracies in the forcing fields (section2) and partly due to deficiencies in the model’s thermal struc-ture. For example, the cooling in reality may have been en-hanced by the reduced solar heating in the east caused by the

smoke from Indonesian fires during this period [Murtugudde etal., 1999], which has not been accounted for in our simulation.

Plate 2b illustrates the horizontal structure of SST (color)and surface current (vectors) anomalies for the control run.Prominent features are cold SSTAs throughout the eastern IOon and south of the equator and a v-shaped pattern of warmSSTAs with its apex in the western IO and branches extendingnortheastward and southeastward. The eastern cooling isstrongest at the eastern boundary and appears to be advectedwestward along the equator via the anomalous currents (aproperty that is verified in section 3.2).

3.1.2. Vertical structure. Plate 3 shows equatorial sec-tions of model temperatures and currents averaged from Oc-tober through December (OND) for the control run (Plate 3b)and for the climatological solution (Plate 3a). The solid curvesdesignate the bottom of the mixed layer. The thermal structureis strikingly different between the two solutions. In the clima-tological case the thermocline slopes down from west to east,attaining a depth of ;100 m at the eastern boundary. Themixed layer thickness, however, remains at ;50 m, so thatthere is a significant barrier layer in the east [Lukas and Lind-strom, 1991]. In contrast, for 1997 the situation is reversed bythe anomalous easterlies. In this case the thermocline slopesupward from west to east, and there is a significant barrierlayer in the west. The mixed layer is ;35 m deep in the westand has reached a minimum specified thickness of 10 m in theeast.

In the climatological solution there is an eastward current inthe upper 50 m almost everywhere, the exception being in theSomali Current region near the west coast. This flow, the

Plate 3. Equatorial sections averaged from October through December showing temperatures (color) andzonal and vertical current vectors for (a) the climatological simulation and (b) the control run for 1997. Solidlines indicate the bottom of the mixed layer.

MURTUGUDDE ET AL.: 1997–1998 INDIAN OCEAN ANOMALOUS EVENTS3298

solution’s fall Wyrtki Jet, is driven by westerly winds that typ-ically peak in November. The flow in the thermocline is weaklyeastward, and there is little indication of upwelling into themixed layer. In contrast, during the fall of 1997 the near-surface flow is completely reversed, the eastward flow in thethermocline is much stronger, and there is upwelling into themixed layer almost everywhere along the section. Thus thevertical structure during fall 1997 is quite similar to those in theequatorial Pacific and Atlantic Oceans, in which a surfaceSouth Equatorial Current (SEC) and a subsurface EquatorialUndercurrent (EUC) are forced by equatorial easterly winds.

3.2. Processes

We utilize the model’s mixed layer temperature equation todefine quantitative measures of the process anomalies thatdetermine SSTAs. That equation is

T1t 5 Q/h1 2 u1T1x 2 v1T1y 1 w11~T2 2 T1!/h1 1 kt¹

2T1,(1)

where T1, h1, and v1 are mixed layer temperature, thickness,and velocity, respectively, Q is the surface heat flux, w1

1 is theentrainment velocity, and kt is the Laplacian mixing coeffi-cient. Each term on the right-hand of (1) measures the influ-ence of a particular process, namely (from left to right), surfaceheating 4, zonal advection 8, meridional advection 9, entrain-ment cooling 0, and mixing; the term on the left-hand sidemeasures temperature tendency 7. Areal averages of theanomalies of individual terms then provide measures of theprocesses governing the temperature variability. In the follow-ing we report anomalies of 4, 8, 9, and 0 in order to showtheir relative importance in determining SSTA. Anomalies arecomputed for each month by subtracting their climatologicalvalues computed over 1982–1998.

3.2.1. Eastern cooling. Figure 1 shows model and Reyn-olds SSTAs in three areas of the eastern IO: along the equatorin the interior ocean (758–858E, 58S–58N), in the eastern equa-torial region adjacent to the coast of Sumatra (858–958E, 58S–58N), and in the coastal upwelling region off Java (958–1058E,108S–08N). In each region, model SSTAs differ from ReynoldsSSTAs by ,0.58C. These differences are most likely related tothe fact that we use climatological radiation and cloudiness. Itis also possible that the model pycnocline shoaled excessivelybecause of strong alongshore winds, thereby strengtheningcooling due to entrainment in the solution. Note that the SSTAvariations in all three regions have similar structures, withperiods of summertime warming followed by cooling duringthe fall and warming during the winter. They differ, however,in the strengths of the warming and cooling events, the sum-mertime warming being strongest along the equator (Figure1a) and the cooling strongest in the coastal upwelling region(Figure 1c).

To identify the causes of these anomalies, Figure 2 showscurves of process anomalies in the three regions. In each re-gion the wintertime warming is caused by Q being large andpositive, the other terms all being negative at that time. In thecoastal upwelling region the increase in Q is largely a reactionto the drop in SST, which lowers the latent heat loss from theocean. In the other two regions, when SSTAs are significantlypositive, they are caused by weakened latent heat loss due toanomalously low wind speeds (see Plate 5). The winter warm-ing in the coastal upwelling region was also facilitated by theshallowing of the mixed layer by as much as 20 m, whichallowed the SST to warm particularly rapidly.

The fall cooling has different causes in each region. Asexpected, in the two coastal regions (Figures 2b and 2c), en-trainment cooling 0 is a prominent cause. Off Java (Figure2c), meridional advection anomalies 9 also contribute signifi-cantly: cold SSTAs appear first off the Java coast in the south-east, and they are subsequently advected northwestward alongthe coast. Note that during normal years the fall season isdominated by the Wyrtki Jet and the advection is southeast-ward. In contrast, in the interior equatorial region (Figure 2a),westward advection of cold waters 8 by the anomalous west-ward jet was the dominant cooling process, as is evident by theequatorial tongue of cool SSTAs in Plate 2. The cooling in bothequatorial regions (Figures 2a and 2b) is also enhanced by aperiod of negative Q during the summer and fall. It is causedby the increased easterlies, which intensify the latent heat lossfrom the ocean, and it lasts until SSTAs decrease, therebyweakening the heat loss.

The large summertime warming in the interior equatorialregion results from positive Q , 0, and 9 anomalies that coun-teract the cooling effect of 8. The initial increase in Q is dueto weaker winds and reduced latent heat loss, and it continuesuntil the increased SST and stronger easterlies enhance latent

Figure 1. Model (solid) and Reynolds (dashed) SST anom-alies averaged over (a) 758–858E, 58S–58N; (b) 858–958E, 58S–58N; and (c) 958–1058E, 108S–08N.

3299MURTUGUDDE ET AL.: 1997–1998 INDIAN OCEAN ANOMALOUS EVENTS

cooling. The largest warming process, however, is due to 9.The meridional velocity in this region (758–858E, 58S–58N)increased from near zero in January 1997 to nearly 3 cm s21 byMay on both sides of the equator and remained positive tillJuly 1997. SST anomalies increased from negative to the southof 28S to positive to the north of the equator till July. ByAugust, positive SSTA occupied the entire box. This led to arapid increase in 9 till August 1997, as seen in Figure 2a.Zonal current anomalies were negative from January throughJuly 1997. Thus positive 9 was being compensated by 8 in theeastern equatorial box. As the easterlies strengthened andeastern cooling ensued, the SSTs and 9 began to drop fromSeptember 1997.

The preceding analysis of terms in (1) is not able to assessthe relative importance of equatorial versus alongshore wind

anomalies in driving the upwelling off Sumatra. To do this, weperformed test runs in which the interannual variability of thewinds was separately suppressed both east and west of 908E. Inthe test without interannual forcing east of 908E the maximumcooling off Sumatra decreased by 40% from its value in thecontrol run, whereas in the other test it decreased by 25%.These results indicate that both local and remote anomalousforcing enhanced the Sumatran upwelling roughly equally. Theanomalous equatorial easterlies generate equatorial andcoastal Kelvin waves that raise the thermocline and thin thebarrier layer off Sumatra (as in Plate 3); hence the cooling dueto upwelling driven by the local alongshore winds is enhancedbecause cooler water is entrained into the mixed layer. TheTOPEX data for fall 1997 clearly show upwelling equatorialKelvin waves [Murtugudde et al., 1998b].

3.2.2. Western and southern warming. To analyze thewarming during the event, we selected three representativeregions: a western equatorial region (458–558E, 58S–58N),which underwent the maximum warming, and the western(608–708E, 158–58S) and eastern (858–958E, 158–58S) parts ofthe southern tropical Indian Ocean. Figures 3 and 4 show

Figure 2. Anomalies of net heat flux, entrainment, and zonaland meridional advection for the mixed layer averaged over (a)758–858E, 58S–58N; (b) 858–958E, 58S–58N; and (c) 958–1058E,108S–08N. All three eastern regions are dominated by entrain-ment cooling during fall 1997. Weaker winds and reducedlatent heat loss lead to rapid increase in net surface heat fluxand warm SST anomalies by early 1998.

Figure 3. Model (solid) and Reynolds (dashed) SST anom-alies averaged over (a) 458–558E, 58S–58N; (b) 608–708E, 158–58S; and (c) 858–958E, 158–58S. Warming of the southern trop-ical Indian Ocean (Figures 3b and 3c) in the model is delayedby over a month compared to in Reynolds SST.

MURTUGUDDE ET AL.: 1997–1998 INDIAN OCEAN ANOMALOUS EVENTS3300

SSTAs and the process anomalies in each of the regions. Wehave included the southeastern region (858–958E, 158–58S)here although it experienced mild cooling during fall 1997 andwarmed only slightly during early 1998.

3.2.2.1. Equatorial region: In the equatorial region, SSTAsrose steadily throughout the event (Figure 3a). This rise isgenerally consistent with the SSTA response determined fromReynolds SST, except that the latter remained nearly constantfor July–October 1997 and then rapidly increased from No-vember 1997 to February 1998.

The corresponding process anomalies are shown in Figure4a. Interestingly, throughout the record, Q only provides cool-ing, a consequence of increased latent cooling due to the warmSSTAs. The 9 and 8 anomalies contributed to the initialwarming. Between May and September, anomalous 9 and 8were nearly equal and opposite in sign, and so, the anomalouswarming was due primarily to 0. The reduced rate of warming

in July and August coincides with a deepening of the mixedlayer and increased entrainment cooling that lowered 0. Notethat there are no large warming terms during the fall andwinter months despite the fact that the SSTA curve continuedto increase steadily. During this period then a significant con-tributor to the temperature balance is the horizontal mixingterm in (1).

The contribution of waves to SSTA cannot be determinedfrom our analysis of terms in (1). Both model and TOPEX/Poseidon data, however, show that a downwelling Rossby wavearrived at the western coast in the equatorial region duringDecember 1997 (Plate 4). This Rossby wave was actually partof a packet that was centered on 88S [Murtugudde et al., 1998b].It warmed the western equatorial IO primarily by deepeningthe pycnocline, both directly and indirectly via reflection ofcoastal and equatorial Kelvin waves from the western bound-ary, thereby decreasing entrainment cooling. This effect isshown in Figure 4a by the rise in 0. Model sea level anomalies,however, are nearly 40% weaker than TOPEX/Poseidon data,and there are also differences in the Rossby wave phase speed,as seen in Plate 4. Thus the model probably fails to produce asufficiently deep thermocline during late 1997 and so under-estimates the role of the decreased entrainment cooling duringJanuary–February 1998.

3.2.2.2. Southern region: The southern tropical IO alsowarmed by over 18C (Figure 3b) in the west and by over 0.58Cin the east (Figure 3c) during the late fall and winter in boththe model and the observations. In the western region (Figure3b), though, warming in the model occurs after October 1997,more than 2 months later than warming in the Reynolds SST.It is not clear what accounts for this discrepancy. Note that themodel SSTAs are colder than Reynolds in both the southerntropical boxes. It is possible that lack of the IndonesianThroughflow leads to a shallower thermocline and hence in-tensified entrainment cooling in the model, resulting in thesolution being too cold throughout the summer. Earlier in therecord, observed and modeled SSTAs also disagree in theeastern region (Figure 3c), but that is to be expected since theamplitudes of the SSTAs there are so small (,0.38C).

Interestingly, the causes of the fall and winter warmings aredifferent in the two regions. In the eastern region (Figure 4c)it is almost entirely due to increased Q , which is caused byweakening winds and reduced latent heat loss. In the westernregion (Figure 4b) the warming results from Q , 0, and 9.Reduced 0 dominates the west (Figure 4b) from June throughNovember 1997, which is, again, related to weaker winds andassociated mixed layer changes. The 8 anomalies generateslightly negative SSTAs in the model until November. At theselatitudes the Indonesian Throughflow has the largest effect onzonal currents, which is not accounted for in our simulations.The warming in the model after November is caused by areduction in 8, enhanced warming due to 9, and increased 4.The meridional velocity was positive between May and No-vember 1997 in the western box and between January and July1997 in the eastern box, explaining the patterns of 9 in Figures4b and 4c. Latent heat loss and 0 start to increase fromDecember onward because of the increased southeasterliesseen in Plate 5, but 9 continues to dominate, and it sustainssurface warming into February 1998.

The cool SSTAs also have different causes in each region. Inthe western region (Figure 4b), they are due entirely to zonaladvection of cooler waters from the eastern ocean, the otherterms tending to warm the ocean. In the eastern region (Figure

Figure 4. Anomalies of net surface heat flux, entrainment,and zonal and meridional advection terms averaged over (a)458–558E, 58S–58N; (b) 608–708E, 158–58S; and (c) 858–958E,158–58S. Figure 4a also shows the anomaly of the rate of SSTchange (see (1)).

3301MURTUGUDDE ET AL.: 1997–1998 INDIAN OCEAN ANOMALOUS EVENTS

Plate 4. (a) Model and (b) TOPEX sea level anomalies along the equator for September 1997 throughJanuary 1998. Note that the color scales are different for Plates 4a and 4b.

Plate 5. Anomalies of wind-speeds (colors; m s21) and wind stresses (vectors, dynes cm22) for January–February 1998.

MURTUGUDDE ET AL.: 1997–1998 INDIAN OCEAN ANOMALOUS EVENTS3302

4c) they are caused by strengthened entrainment 0 through-out the summer and are sustained (and strengthened) duringthe fall by zonal 8 and meridional 9 advection. The coolingends when these terms begin to weaken in November as thedriving winds reduce in strength. Thus these three boxes wereall within the v-shaped warm SST region, but the warming wasostensibly caused by different oceanic processes.

3.3. Other Events

3.3.1. Time series, correlations, and EOFs. Plate 6 showsSSTAs from the control run for the entire time span of theintegration, averaged over the eastern (908–1008E, 58N–58S)and western (458–558E, 58N–58S) equatorial IO. It is clear thatneither the eastern cooling nor the western warming is uniqueto 1997. There are two other extreme events in the record inwhich the cooling was .218C in 1961 and 1994, and a numberof events with anomalies of the order of 20.58C. Moreover,significant warm episodes in the western IO tend to occurtogether with the strong cool eastern events, the correlationbetween the eastern and western time series in Plate 6 being20.44, which is statistically significant at 90% confidence level.Plate 7 plots SSTAs in the eastern equatorial IO (averagedover 908–1008E, 58N–58S) together with anomalies of equato-rial zonal wind stress (658–858E, 58S–58N) and meridionalwinds off Sumatra (858–1008E, 108S–08N). The correlationsbetween SSTA and winds are very high (0.83 and 20.66, re-spectively). Finally, the correlation between equatorial zonalwind stress and thermocline depth anomalies averaged in theeastern equatorial region is 0.7, an indication that remoteforcing by equatorial Kelvin waves is acting in all events inaddition to local Ekman pumping.

These significant correlations suggest that anomalous SSTAand wind fields associated with all IO events have similar

structures. To confirm this idea, we analyzed the first EOFs forSSTA and wind stresses for the entire period of the control run(not shown). Not surprisingly, the structures of the EOFs werestrikingly similar to the SSTA and wind anomaly patterns inPlate 2. To ensure that the wind pattern is really associatedwith the dominant mode of SST variability, we carried outmultivariate EOF analyses for SST and wind stresses. The firstmode of the combined EOF is shown in Plate 8, which clearlyconfirms the fact that model SSTAs are coupled to the windpattern for all such events.

3.3.2. Natural mode of oscillation. The similar structureof all IO events suggests that they are a natural mode ofoscillation in the IO, easily excited in a variety of ways. One ofthose ways is forcing by climatic events external to the IO, theprime example of which is ENSO. Indeed, note in Plate 6 thatmany of the stronger coolings occurred during ENSO years(e.g., 1972–1973, 1982–1983, 1986–1987, 1991–1992, and 1997–1998). Moreover, 1994 was arguably a moderate ENSO year,and there was a severe Indonesian drought in 1961, usually acertain indicator of ENSO. (Apparently, 1961 was an instancewhere ENSO-like conditions prevailed only in the westernbasin [Flohn, 1986].) The relationship with ENSO is evidencedby significant correlations with the Southern Oscillation Index(SOI), with correlations between zonal and meridional windanomalies and the SOI attaining values of 0.57 and 20.51,respectively, in the region 08N–108S, 908–1058E. A likely causallink between the IO events and ENSO is via the atmosphere[Venzke et al., 1997].

On the other hand, the SOI correlations are not that largebecause many of the events do not occur with ENSO, and evenif they do, they do not occur with the same amplitude. Latif andBarnett [1995] carried out several uncoupled and hybrid-coupled model experiments to demonstrate that while the air-

Plate 6. Time plots of SST Anomalies for 1958–1997. Extreme cold anomalies of more than 18C occur in1961, 1994 and 1997. Weaker events occur regularly throughout the record.

3303MURTUGUDDE ET AL.: 1997–1998 INDIAN OCEAN ANOMALOUS EVENTS

sea interactions in the IO are weaker than those in the Pacific,they do contribute significantly to the local SST variability.Thus the ENSO teleconnections are often enhanced by localair-sea interactions. These properties suggest the possibilitythat anomalous events are also amplified by, or even resultfrom, ocean-atmosphere interactions internal to the IO. Con-sider the following scenario. Suppose that cool SSTAs appearin the eastern ocean, where they initiate an eastward shift ofIndonesian convection. This shift allows organized convectionto develop in the western ocean, establishes equatorial easter-lies, sets off the oceanic events discussed in this paper, andhence leads to an intensification of the eastern cooling. Such apositive feedback mechanism (essentially the same one pro-posed by Bjerknes [1969] to explain El Nino) has been sug-

gested as an explanation for the 1961 rain event [Reverdin et al.,1986; Kapala et al., 1994]. Reverdin et al. [1986] also point outthat negative (positive) cloud anomalies associated with cold(warm) SSTA during such events most likely imply coupledinteractions.

4. Summary and ConclusionsIn our paper we have used an ocean model to examine the

structure of interannual variability in the IO from 1958–1997,focussing on the intense 1997–1998 event. The model simula-tion captures the eastern equatorial and coastal cooling duringOctober–December 1997 and the rapid warming by January1998. The model warmings in the western equatorial region

Plate 7. Correlations of SST and wind stress anomalies over (a) 58S–58N, 908–1008E and 58S–58N, 658–858E,respectively, and (b) 58S–58N, 908–1008E and 108S–08N, 858–1008E, respectively. SST anomalies over easternequatorial Indian Ocean show a very high correlation with zonal wind anomalies over central equatorialregion, indicating that remotely forced Kelvin waves affect eastern coastal regions.

MURTUGUDDE ET AL.: 1997–1998 INDIAN OCEAN ANOMALOUS EVENTS3304

and the southern tropical IO are also in good agreement withReynolds SSTA.

It is shown that equatorial easterly wind stress anomaliesproduced eastward Kelvin waves that raised the thermoclineoff the Sumatran coast. Strengthened alongshore winds en-hanced surface cooling by driving stronger upwelling there,which persisted till the end of 1997. The rapid collapse of thewinds, the resultant reduction in latent heat losses, and theshallowing of the mixed layer led to elimination of the coldSSTAs by February 1998. Away from the coast in the easternequatorial region (758–858E, 58S–58N), advection played animportant role in both the warming during May–July 1997 andthe cooling during August–November 1997. The coastal re-gions were dominated by entrainment cooling during most of1997. Wind speed dropped to near zero by late 1997 and early1998, which resulted not only in the shallowing of the mixedlayer but also in a reduction in latent heat losses. The negativeSSTAs were replaced by near-zero or slightly positive SSTAsby February 1998 in the eastern part of the basin.

The western equatorial region underwent surface warmingfrom May 1997 onward, with peak SSTAs of nearly 28C byFebruary 1998. A weaker Southwest Monsoon resulted in ini-tial warming due to anomalous meridional advection and re-duced entrainment cooling. The net heat flux anomalies re-mained negative and appeared to be a negative feedback tosurface warming. Horizontal mixing and an incident down-welling Rossby wave seem to have been crucial in the modelfor sustaining the warming during late 1997 and early 1998.The amplitude of the Rossby wave in the model is nearly 40%weaker than in TOPEX/Poseidon data, however, indicatingthat the model thermocline may be shallower than in reality.This discrepancy may have led to an underestimation of therole of local processes in the equatorial warming during Janu-

ary–February 1998. The western southern tropical IO experi-enced a warming from September 1997 onward, with peakSSTA of over 18C in February 1998. Anomalous meridionaladvection and reduced mixed layer entrainment were largelyresponsible for the warming with some contribution from netheat flux during fall. The warming in the model simulationdoes not commence until October 1997, fully 2 months laterthan the observed warming. The eastern part of the southerntropical IO was warmer by ;0.58C during January–February1998, mainly because of weaker winds and reduced latent heatloss.

An empirical orthogonal function (EOF) analysis of themodel and Reynolds SSTA shows that the spatial pattern ofthe first model is remarkably similar to the patterns observedduring 1961, 1994, and 1997. It is thus possible that the anom-alous events during these years represent amplifications of acoupled IO mode. The forces, either internal or external, thatamplify the natural mode remain to be identified, especiallythe processes that initialize what may potentially be positivefeedback from the IO to the atmosphere above it. Within theIO region the connection to ENSO appears to be largelythrough the atmosphere via shifts in the Walker Circulationand associated radiative and wind-forced evaporative changes[Venzke et al., 1997]. There is, however, also an oceanic con-nection through changes in the Indonesian Throughflow trans-port [Meyers, 1996], which have large-scale, dynamic, and ther-modynamic effects on IO circulation [Murtugudde et al.,1998a].

In conclusion, while we have not answered all possible ther-modynamic and dynamic questions, we have analyzed, in de-tail, the oceanic processes that accompany IO events such asthe one observed in 1997–1998. In a purely wind-driven modelsuch as ours it is not possible to determine with certainty if

Plate 8. Combined first empirical orthogonal function (EOF) of the anomalies of model SST (1958–1997;color) and NCEP wind stresses show that the spatial pattern is strikingly similar to the SST anomalies andwind stress anomalies in Plate 2.

3305MURTUGUDDE ET AL.: 1997–1998 INDIAN OCEAN ANOMALOUS EVENTS

positive feedback is active in the IO events. Experiments withcoupled models that are able to simulate the events will berequired to confirm the idea.

Acknowledgments. The authors wish to thank Stuart Godfrey andseveral members of the CLIVAR Monsoon Panel for very insightfuldiscussions. We deeply appreciate the assistance of James Beauchampwith some of the analyses and figures. We are indebted to Jim O’Brienfor allowing us access to the FSU computational facilities. RM waspartially supported by NASA grants RTOP 622-32-51-20 (TRMM)and 621-55-24 (TOPEX Extended Mission) for which he is extremelygrateful. JPM was supported by NSF grant OCE-94-16098, ONR con-tract N00014-97-1-0077, and the Frontier Research System for GlobalChange through its sponsorship of the International Pacific ResearchCenter (IPRC). AJB would like to acknowledge NASA grant 578-21-03.

ReferencesBirkett, C., R. Murtugudde, and T. Allan, Indian Ocean climate event

brings floods to East Africa’s lakes and Sudd Marsh, Geophys. Res.Lett., 26, 1031–1034, 1999.

Bjerknes, J., Atmospheric teleconnections from the equatorial Pacific.Mon. Weather Rev., 97, 163–172, 1969.

Chen, D., L. M. Rothstein, and A. J. Busalacchi, A hybrid verticalmixing scheme and its application to tropical ocean models, J. Phys.Oceanogr., 10, 2156–2179, 1994.

Flohn, H., Indonesian droughts and their teleconnections, Ber. Geogr.Stud., 20, 251–265, 1986.

Kalnay, E., et al., The NCEP/NCAR 40 year reanalysis project, Bull.Am. Meteorol. Soc., 77, 437–471, 1996.

Kapala A., K. Born, and H. Flohn, Monsoon anomaly or an El Ninoevent at the equatorial Indian Ocean? Catastrophic rains 1961/62 inEast Africa and their teleconnections, paper presented at Proceed-ings of the International Conference on Monsoon Variability and Pre-diction, World Meteorol. Org., Trieste, Italy, 1994.

Latif, M., and T. Barnett, Interactions of the tropical oceans, J. Clim.,8, 952–964, 1995.

Levitus, S., Climatological atlas of the world ocean, NOAA Prof. Pap.13, U.S. Govt. Print. Off., Washington, D. C., 1994.

Lukas, R., and E. Lindstrom, The mixed layer of the western equato-rial Pacific, J. Geophys. Res., 96, 3343–3357, 1991.

Meyers, G., Variation of Indonesian Throughflow and the El Nino-Southern Oscillation, J. Geophys. Res., 101, 12,255–12,263, 1996.

Murtugudde, R., and A. J. Busalacchi, Salinity effects in a tropicalocean model, J. Geophys. Res., 103, 3283–3300, 1998.

Murtugudde, R., and A. J. Busalacchi, Interannual variability of thedynamics and thermodynamics of the tropical Indian Ocean,J. Clim., 12, 2300–2326, 1999.

Murtugudde, R., R. Seager, and A. Busalacchi, Simulation of theTropical Oceans with an ocean GCM coupled to an atmosphericmixed layer model, J. Clim., 9, 1795–1815, 1996.

Murtugudde, R., A. J. Busalacchi, and J. Beauchamp, Seasonal tointerannual effects of the Indonesian Throughflow on the tropicalIndo-Pacific basin, J. Geophys. Res., 103, 21,425–21,441, 1998a.

Murtugudde, R., A. J. Busalacchi, and J. Beauchamp, IndonesianThroughflow and ENSO: Indo-Pacific waves from TOPEX/Poseidon, paper presented at Joint TOPEX/Poseidon Jason-1 Sci-ence Working Team Meeting, NASA, Keystone, Colo., 1998b.

Murtugudde, R., S. Signorini, J. Christian, A. Busalacchi, C. McClain,and J. Picaut, Ocean color variability of the tropical Indo-Pacificbasin observed by SeaWiFS during 1997–1998, J. Geophys. Res., 104,18,351–18,366, 1999.

Reverdin G., D. Cadet, and D. Gutzler, Interannual displacements ofconvection and surface circulation over the equatorial Indian Ocean,Q. J. R. Meteorol. Soc., 112, 43–67, 1986.

Venzke, S., M. Latif, and A. Vilwock, The coupled GCM ECHO-2, II,Indian Ocean response, Rep. 246, Max-Planck-Inst. fur Meteorol.,Hamburg, Germany, 1997.

Xie, P., and P. Arkin, An intercomparison of gauge observations andsatellite estimates of monthly precipitation, J. Appl. Meteorol., 34,1143–1160, 1996.

A. J. Busalacchi, Laboratory for Hydrospheric Processes, NASAGoddard Space Flight Center, Greenbelt, MD.

J. P. McCreary Jr., International Pacific Research Center, Universityof Hawaii, Honolulu, HI.

R. Murtugudde, ESSIC, University of Maryland, College Park, MD20742. (ragu@barolo.gsfc.nasa.gov)

(Received April 16, 1999; revised September 30, 1999;accepted October 25, 1999.)

MURTUGUDDE ET AL.: 1997–1998 INDIAN OCEAN ANOMALOUS EVENTS3306